MULTI-SPECTRAL IMAGE FUSION METHOD BASED ON WAVELET 
TRANSFORMATION 
a YAO Wan-qiang b ZHANG Chun-sheng 
d Dept, of Survey Engineering, Xi’an University of Science & Technology, Xi’an 710054, China, - sxywq@163.com 
b Dept, of Survey Engineering, Xi’an University of Science & Technology, Xi’an 710054, China - chshzh@sohu.com 
Commission VII, WG Vll/6 
KEY WORDS: Image fusion; Wavelet transform; Weighting average; Threshold; Characteristic of human vision system 
ABSTRACT: 
The paper focuses on image fusion between multi-spectral images and panchromatic images using a wavelet analysis method with 
good signal processing and image processing traits. A new weighting technique is developed based on wavelet transformation for the 
fusion of a high spatial resolution image and a low-resolution, multi-spectral image. The method improves a standard wavelet 
merger for merging the lower frequency components of a multi-spectral image and its high spatial resolution image by means of 
local deviation rules with weighting average. And then the merged image is reconstructed by an inverse wavelet transform using the 
fused approximation and details from the high spatial resolution image. Also, a multi-spectral images fusion algorithm is proposed 
based on wavelet transform characteristic of human vision system. Firstly, perform a wavelet multi-scale transformation of each 
source image. Then a new fusion regular is presented based on human vision system corresponding high (low) frequency 
components are divided into several blocks, and contrast error of every block is calculated, an adaptive threshold selection is 
proposed to decide which should be used to construct the new high (low) frequency components. Finally, the fused image is 
obtained by taking inverse wavelet transform. The experimental results show that the new method presented is clearly better in not 
only preserving spectral and improving spatial presentation, but also avoiding mosaic occurring. 
1. INTRODUCTION 
2. IMAGE FUSION BASIC FLOW 
The image fusion is that the multiple images which obtains 
from a sensor or many sensors synthesizes an image, in which 
the information from the multiple primitive images can be 
reflected so as to analyze and judge the target more precisely 
and comprehensively. Because both the images gain from 
multi-sensors have the redundancy and the complement, the 
multi-sensor image fusion technology may enhance the 
reliability of the system and also enhance the use efficiency of 
the pictorial information [1]. At present, various militarily 
significant states in the world competitively invest massive 
manpower, physical resource and financial resource to carry 
on the information fusion technology and have obtained 
magnificent research results. Take US for example, the 
expense that is used, every year, in the research of 
information fusion technology amount to more than 
100,000,000 US dollar. The image fusion technology in such 
aspects as medicine, remote sensing, computer vision, weather 
forecast has also been widely applied. Especially in the 
computer visual aspect, in the astronautics and aviation multi 
delivery platform, the massive remote sensing image fusion 
obtained from each kind of remote sensor in different spectra, 
different wave bands, different temporal or different angles 
provides good processing method for information highly 
effective extraction, and obtains obvious benefit. In the last 
few years, along with the information fusion technology 
development, obtaining the remote sensing image in double 
high resolution with the post-processing method has become 
the essential target and the duty of the remote sensing 
information fusion, and has formed many algorithms, like IHS 
algorithm, PCA algorithm and that based on wavelet 
transformation algorithm. 
The commonly classification of image fusion is based on the 
image attribute, which divides image fusion into three levels, 
namely picture element level, characteristic level and policy 
making level fusion. The object of image fusion data may can 
be divided into the optical image and the non-optical image 
according to the image formation way. The basic processes of 
Figure 1 Basic procedure of image fusion between multi- 
spectral images and panchromatic images